System Detail
iic/eres2netv2-zh
eresv2_zh
Submitted by: Public checkpoint
- Global position
- #11 / 23
- Scenario-macro EER
- 11.5449
- Scenario-macro minDCF
- 0.484961
- Trial-set coverage
- 26 / 26
- Scenario top-three
- 1
Publicly ranked
This result is included in the reviewed public ranking.
Model Provenance
Source and architecture
- Displayed name:
iic/eres2netv2-zh - Submitted by: Public checkpoint
- Metadata: OpenSVBench model record
- Source: IIC
- Architecture: ERes2NetV2
- Model family: ERes2Net
- Submitted alias:
speech_eres2netv2_sv_zh-cn_16k-common - Created at:
2026-05-29T14:35:20.759435+00:00
Training And Links
Training data and references
- Training data: Large Chinese speaker corpus, about 200k speakers
- Training setup: Official 16 kHz Chinese ERes2NetV2 release.
- Submission mode:
full-core - Model source: Open link
- Paper: ERes2NetV2: Boosting Short-Duration Speaker Verification Performance with Computational Efficiency
Higher-Ranked Scenarios
Scenario ranks above the global position
Show 3 supporting trial sets
- CN-Celeb: #1 on its trial-set ranking.
- CN-Celeb Genre: #1 on its trial-set ranking.
- GSC Short: #3 on its trial-set ranking.
Lower-Ranked Scenarios
Scenario ranks below the global position
Show 3 supporting trial sets
- VoxCeleb Short: #20 on its trial-set ranking.
- VoxCeleb1-H: #19 on its trial-set ranking.
- VoxCeleb1-O: #19 on its trial-set ranking.
Scenario Rankings
Full ranking across scenarios
This table shows where the model sits on each scenario, using source-balanced scenario scores and the linked trial sets as evidence.
| Scenario | Rank | Lens | Evidence | Score |
|---|---|---|---|---|
|
Genre-Shift Robustness
Whether performance holds when CN-Celeb enrollment and test speech come from different source genres.
|
#1/23
Leader
|
Source-genre variation
0.0000 EER from leader
|
14.2725
0.527525 minDCF
|
|
|
Channel / Device Robustness
How resilient the model is to device and channel mismatch.
|
#7/23
Competitive
|
Device and channel variation
7.9735 EER from leader
|
11.9758
0.644862 minDCF
|
|
|
Short-Duration Robustness
How much performance holds up when speech evidence is limited by duration.
|
#7/23
Competitive
|
Short-duration speech
3.9349 EER from leader
|
12.9695
0.460887 minDCF
|
|
|
In-The-Wild Robustness
How strong the model is on unconstrained celebrity and media speech across official CN-Celeb and VoxCeleb protocols.
|
#8/23
Competitive
|
Open-domain media speech
1.8152 EER from leader
|
5.5369
0.294964 minDCF
|
|
|
Distance Robustness
How much performance changes across meeting and domestic distance mismatch.
|
#10/23
Competitive
|
Distance mismatch
6.0169 EER from leader
|
11.2737
0.537866 minDCF
|
|
|
Cross-Lingual Robustness
How well speaker identity survives enrollment-test language mismatch.
|
#11/23
Competitive
|
Language mismatch
3.6743 EER from leader
|
5.3320
0.299692 minDCF
|
|
|
Speaking-Style Robustness
Whether the model can preserve identity across emotion-driven change, whispered speech, and noise-induced Lombard speaking style.
|
#13/23
Needs work
|
Speaking style shift
4.0521 EER from leader
|
5.9688
0.361679 minDCF
|
|
|
Accent / Dialect Robustness
How reliably the model tracks identity across accent and dialect mismatch.
|
#15/23
Needs work
|
Accent and dialect variation
6.9967 EER from leader
|
23.8183
0.764092 minDCF
|
|
|
Aging Robustness
How stable identity representations remain across age-derived and longitudinal recording gaps.
|
#17/23
Needs work
|
Speaker aging and time gaps
7.4947 EER from leader
|
9.2864
0.497259 minDCF
|
|
|
Overlap Robustness
How well speaker identity survives light, mid, and heavy overlap in both meeting and domestic recordings.
|
#17/23
Needs work
|
Overlapping speakers
6.7707 EER from leader
|
14.8800
0.543251 minDCF
|
|
|
Noise / Reverb Robustness
How reliably the model preserves identity when room acoustics, distractor noise, and microphone placement deviate from an easier in-corpus reference condition.
|
#18/23
Needs work
|
Noise and reverberation
9.4200 EER from leader
|
11.6800
0.402500 minDCF
|
Variant Compare
Compare ERes2Net variants
Expand this section to compare models in the same model family by architecture, source, training data, training setup, and leaderboard result.
Show
Compare ERes2Net variants
Expand this section to compare models in the same model family by architecture, source, training data, training setup, and leaderboard result.
| Variant | Submitted by | Source | Training Data | Training Setup | Global Rank | Macro EER | Macro minDCF | Open |
|---|---|---|---|---|---|---|---|---|
|
iic/eres2netv2-zh
ERes2NetV2
Current
|
Public checkpoint | IIC | Large Chinese speaker corpus, about 200k speakers | Official 16 kHz Chinese ERes2NetV2 release. |
#11
ranked
|
11.5449 | 0.484961 | Open |
|
iic/eres2net-en
ERes2Net
|
Public checkpoint | IIC | VoxCeleb2 development set, 5,994 speakers | Official 16 kHz English ERes2Net release. |
#13
ranked
|
12.0955 | 0.472802 | Open |
|
iic/eres2net-large-3dspeaker
ERes2Net
|
Public checkpoint | IIC / ModelScope | 3D-Speaker open dataset, about 10k Chinese speakers | Official 16 kHz Chinese ERes2Net-Large release; the local model card reports 18.3M parameters and training on 3D-Speaker. |
#17
ranked
|
14.1682 | 0.593516 | Open |
Trial-Set Results
Raw ranking by trial set
Expand this section to inspect the detailed trial-set rankings.
Show
Raw ranking by trial set
Expand this section to inspect the detailed trial-set rankings.
| Trial Set | Rank | Scenarios | Trials | Score |
|---|---|---|---|---|
|
TidyVoiceX2-ASV
TidyVoiceX2-ASV
|
#11/23
|
Cross-lingual
|
200000 |
5.3320
0.299692 minDCF
|
|
HI-MIA
HI-MIA
|
#4/23
|
Short-duration
|
660000 |
3.7090
0.208377 minDCF
|
|
GSC Short
Google Speech Commands
|
#3/23
|
Short-duration
|
220000 |
5.0750
0.303175 minDCF
|
|
GLOBE
GLOBE
|
#18/23
|
Accent/dialect
|
56848 |
26.7067
0.580650 minDCF
|
|
CN-Celeb
CN-Celeb
|
#1/23
|
In-the-wild
|
3484292 |
3.8243
0.165880 minDCF
|
|
CN-Celeb Genre
CN-Celeb
|
#1/23
|
Genre shift
|
440000 |
14.2725
0.527525 minDCF
|
|
CN-Celeb Short
CN-Celeb
|
#3/23
|
Short-duration
|
546964 |
17.6675
0.559834 minDCF
|
|
VoxCeleb1-O
VoxCeleb1
|
#19/23
|
In-the-wild
|
37611 |
6.3852
0.421188 minDCF
|
|
VoxCeleb1-E
VoxCeleb1
|
#19/23
|
In-the-wild
|
579818 |
6.3659
0.387641 minDCF
|
|
VoxCeleb1-H
VoxCeleb1
|
#19/23
|
In-the-wild
|
550894 |
8.9975
0.463313 minDCF
|
|
VoxCeleb Short
VoxCeleb1
|
#20/23
|
Short-duration
|
394724 |
25.4264
0.772160 minDCF
|
|
3D-Speaker Device
3D-Speaker
|
#9/23
|
Channel/device
|
180000 |
17.4433
0.869280 minDCF
|
|
3D-Speaker Distance
3D-Speaker
|
#5/23
|
Distance
|
175163 |
14.1380
0.723966 minDCF
|
|
3D-Speaker Dialect
3D-Speaker
|
#14/23
|
Accent/dialect
|
180000 |
20.9300
0.947533 minDCF
|
|
FFSVC 2022 Cross-Channel
FFSVC 2022
|
#5/23
|
Channel/device
|
72000 |
6.5083
0.420444 minDCF
|
|
FFSVC 2022 Cross-Domain
FFSVC 2022
|
#5/23
|
Distance
|
66546 |
6.6987
0.430696 minDCF
|
|
Whisper40 Whisper
Whisper40
|
#6/23
|
Speaking style
|
17600 |
9.1250
0.663000 minDCF
|
|
Lombard Grid Lombard
Lombard Grid
|
#17/23
|
Speaking style
|
29524 |
1.2034
0.075149 minDCF
|
|
VOiCES Noise/Reverb
VOiCES
|
#18/23
|
Noise/reverb
|
55000 |
11.6800
0.402500 minDCF
|
|
CHiME-6 Domestic Far-Field
CHiME-6
|
#18/23
|
Distance
|
36487 |
18.3901
0.771721 minDCF
|
|
CHiME-6 Overlap
CHiME-6
|
#17/23
|
Overlap
|
39600 |
23.1667
0.743694 minDCF
|
|
ESD
ESD
|
#17/23
|
Speaking style
|
437408 |
7.5779
0.346889 minDCF
|
|
AliMeeting Near/Far
AliMeeting
|
#4/23
|
Distance
|
220000 |
5.8680
0.225080 minDCF
|
|
AliMeeting Overlap
AliMeeting
|
#6/23
|
Overlap
|
165000 |
6.5933
0.342807 minDCF
|
|
VoxKnesset
VoxKnesset
|
#17/23
|
Aging
|
158312 |
11.4067
0.527584 minDCF
|
|
VoxPopuli Aging
VoxPopuli
|
#17/23
|
Aging
|
146575 |
7.1662
0.466934 minDCF
|